CN101216890A - A color image segmentation method - Google Patents
A color image segmentation method Download PDFInfo
- Publication number
- CN101216890A CN101216890A CNA2008100558332A CN200810055833A CN101216890A CN 101216890 A CN101216890 A CN 101216890A CN A2008100558332 A CNA2008100558332 A CN A2008100558332A CN 200810055833 A CN200810055833 A CN 200810055833A CN 101216890 A CN101216890 A CN 101216890A
- Authority
- CN
- China
- Prior art keywords
- degree
- membership
- matrix
- registration
- matrixes
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000003709 image segmentation Methods 0.000 title claims abstract description 41
- 238000000034 method Methods 0.000 title claims abstract description 28
- 239000011159 matrix material Substances 0.000 claims abstract description 99
- 230000004927 fusion Effects 0.000 claims description 15
- 238000010586 diagram Methods 0.000 claims description 4
- 230000008676 import Effects 0.000 claims description 3
- 230000011218 segmentation Effects 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 4
- 230000000977 initiatory effect Effects 0.000 description 4
- 239000013598 vector Substances 0.000 description 3
- 238000005192 partition Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000003706 image smoothing Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 238000002203 pretreatment Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention relates to a color image segmentation method, which comprises the following steps: an unsegmented image is input and is initialized to set the number of categories of the unsegmented image; a plurality of different fuzzy clustering algorithms are used to respectively segment the image and obtain a plurality of membership grade matrixes correspondingly; one of the membership grade matrixes is taken as a benchmark matrix, the other membership grade matrixes are respectively registrated with category labels of all the membership grade matrixes, and all the membership grade matrixes are re-registrated and arranged according to the registrated category labels, so as to obtain registrated membership grade matrixes correspondingly; the registrated membership grade matrixes and the benchmark matrix are fused to obtain a sued membership grade matrix, and the category label corresponding to each pixel point is calculated according to the fused membership grade matrix, so as to realize the image segmentation. The invention utilizes a plurality of fuzzy clustering algorithms to segment the color image, which can effectively improve the precision of the image segmentation.
Description
Technical field
The present invention relates to a kind of image partition method, especially relate to a kind of color image segmentation method that merges based on the fuzzy clustering algorithm.
Background technology
Along with development of computer, color digital image is increasingly extensive in the application of all trades and professions.The segmented extraction of coloured image is the basis that obtains the important channel of image information and carry out image understanding, also is the major issue of Image Engineering technology.And fuzzy clustering is classical mode identification technology, can effectively solve a lot of artificial intelligence problems, but present color Image Segmentation seldom adopts Fuzzy clustering techniques, even adopting also just only adopts a kind of fuzzy clustering algorithm to carry out image segmentation, not in conjunction with multiple fuzzy clustering algorithm, therefore, segmentation precision remains further to be improved.
Summary of the invention
Technical matters to be solved of the present invention is to provide a kind of color image segmentation method that merges based on the fuzzy clustering algorithm, utilizes multiple fuzzy clustering algorithm that coloured image is cut apart, to improve the precision of image segmentation.
To achieve these goals, the invention provides a kind of color image segmentation method, comprising: import image to be split and carry out initializing set, set the classification number of image to be split; Use multiple different fuzzy clustering algorithm respectively this image to be cut apart, correspondingly obtain a plurality of degree of membership matrixes; With in these a plurality of degree of membership matrixes one of them as the benchmark matrix, classification mark to other degree of membership matrixes of remainder carries out registration respectively, and according to the classification mark behind the registration each degree of membership matrix is carried out again registration and arrange the corresponding degree of membership matrix that obtains behind registration; Described degree of membership matrix and this benchmark matrix behind registration merged, obtain a degree of membership matrix after the fusion, and go out the pairing classification mark of each picture element, thereby realize image segmentation according to the degree of membership matrix computations after this fusion.
Preferably, the multiple different fuzzy clustering algorithm of described use is cut apart this image, and the step that obtains a plurality of degree of membership matrixes comprises: the initial parameter of setting a kind of fuzzy clustering algorithm; Utilize this fuzzy clustering algorithm that this image is cut apart, obtain a degree of membership matrix; Repeat repeatedly above-mentioned steps, utilize multiple different fuzzy clustering algorithm respectively this image to be cut apart, correspondingly obtain a plurality of degree of membership matrixes.
Preferably, described step of carrying out registration comprises: select in these a plurality of degree of membership matrixes one of them as the benchmark matrix, set its classification mark and with as the reference category mark, wherein, the picture element cluster that belongs to same classification is same classification mark; Other degree of membership matrixes for remainder, set the classification mark of each degree of membership matrix respectively, and calculate the distance of the pairing degree of membership of each reference category mark of any one classification mark and this benchmark matrix in the classification mark of each degree of membership matrix, then to answer positioned in registration be the minimum reference category mark of distance to this any one classification mark; With the classification mark of each the degree of membership matrix in other degree of membership matrixes of remainder all registration be corresponding reference category mark, and with reference to registration results, other degree of membership matrixes of remainder are carried out again registration arrangement, a plurality of degree of membership matrixes behind registration of corresponding acquisition respectively.
Preferably, the described step that merges and calculate each picture element corresponding class mark comprises: collect this benchmark matrix and a plurality of degree of membership matrixes behind registration; Calculate the average mutual information of each matrix in this benchmark matrix and a plurality of degree of membership matrixes behind registration; Calculate the weight of each matrix; This benchmark matrix and a plurality of degree of membership matrixes behind registration are weighted fusion, obtain a weighting degree of membership matrix after the fusion; According to the weighting degree of membership matrix after this fusion, calculate the classification mark of each picture element in the fused image, finish image segmentation.
Preferably, in the described step of carrying out registration, the distance of this degree of membership satisfies:
Wherein, N is the number of pixels of image to be split; K, l represent classification; μ
k (1)(x
i) represent in this benchmark matrix (1) that i picture element belongs to the degree of membership of k class; μ
l (t)(x
i) represent that i the picture element of (t) individual matrix in other remaining degree of membership matrixes belongs to the degree of membership of l class.
Preferably, in the described step of carrying out registration, for other degree of membership matrixes of remainder, the classification mark of any one classification mark behind registration satisfies in the classification mark of each degree of membership matrix:
Preferably, described average mutual information Φ
tSatisfy:
Wherein, M is the number of the fuzzy clustering algorithm of use; N is the number of pixels of image to be split; The classification number of K for setting; μ
j (t)(x
iI the picture element of (t) individual matrix belongs to the degree of membership of j class in the degree of membership matrix behind collected M the registration of) ' expression; μ
j (l)(x
iI the picture element of (l) individual matrix belongs to the degree of membership of j class in the degree of membership matrix behind collected M the registration of) ' expression.
Preferably, the weight of the degree of membership matrix behind each registration satisfies:
T=1 wherein, 2 ..., M; Z is a normalized factor, makes
Preferably, described weighting degree of membership matrix satisfies:
I=1 wherein, 2 ..., N; J=1,2 ..., K; μ
j (t)(x
iI the picture element of (t) individual matrix belongs to the degree of membership of j class in the degree of membership matrix behind collected M the registration of) ' expression.
Preferably, the classification mark of described each picture element satisfies:
Each picture element x in this weighting degree of membership matrix wherein
iPairing weighting degree of membership in the pairing classification j of maximal value be this picture element x
iThe classification mark.
Preferably, also comprise before the image segmentation carrying out: the image of being imported to be split is carried out pre-service, remove the noise spot in the image.After finishing image segmentation, also comprise: described image segmentation result is carried out aftertreatment, remove the very little zone of area in noise spot in the image and/or the combined diagram picture.
The present invention utilizes multiple different fuzzy clustering algorithm that coloured image is cut apart, and segmentation result is carried out weighting fusion behind registration, thereby the segmentation result after obtaining merging is realized cutting apart of coloured image, has improved the precision of image segmentation effectively.
Describe the present invention below in conjunction with the drawings and specific embodiments, but not as a limitation of the invention.
Description of drawings
Fig. 1 is the process flow diagram of color image segmentation method of the present invention;
Fig. 2 is the synoptic diagram of the preferable color image segmentation method that merges based on multiple fuzzy clustering algorithm of the present invention one.
Embodiment
As shown in Figure 1, color image segmentation method of the present invention mainly is earlier by initializing set (step 101), and the classification number of the color digital image that needs are cut apart, the number and the clustering algorithm of clustering algorithm carry out initial setting; Utilize multiple different fuzzy clustering algorithm respectively image to be carried out dividing processing (step 102) then, obtain a plurality of degree of membership matrixes, promptly a plurality of fuzzy clustering algorithms are cut apart the cluster result that obtains to image respectively; Because each cluster result obtains respectively, the order of the classification mark of the degree of membership matrix that obtains may be different, therefore, the classification mark of the degree of membership matrix that each fuzzy clustering algorithm need be obtained carries out registration (step 103), and rearrange, thereby obtain a plurality of degree of membership matrixes behind registration according to registration results; Then, to merging after the degree of membership matrix weighting behind these registrations, promptly merge the cluster result (step 104) behind the registration, obtain a weighting degree of membership matrix after the fusion, can calculate the classification mark of each picture element in this image according to this weighting degree of membership matrix, thereby realize cutting apart image.Preferably, the image after cutting apart also can carry out aftertreatment (step 105), makes segmentation result more reasonable.
Below in conjunction with Fig. 2, good embodiment describes the color image segmentation method that merges based on multiple fuzzy clustering algorithm of the present invention in detail with a religion.
As shown in Figure 2, at first, import image to be split { I (x
i)
I=1 N, total N pixel in this image, I (x
i) be picture element x
iPixel vectors, and carry out initializing set, set the classification number K of image to be split and the M kind fuzzy clustering algorithm that uses.Wherein, described M kind fuzzy clustering algorithm can use the diverse fuzzy clustering algorithm of M kind; Also can use with a kind of fuzzy clustering algorithm, but select different initiation parameters; Or the mixing of both of these case, in a word, as long as this M kind algorithm is not identical.
Preferably, the present invention also can comprise a pre-treatment step after this image of input, can adopt the whole bag of tricks that this image is carried out pre-service, as image smoothing, filtering and noise reduction etc., to remove the noise in the image, carries out image segmentation below being convenient to.
Then, use the different fuzzy clustering algorithm of M kind that the image to be split of input is cut apart, obtain M degree of membership matrix, the result of respectively this image being cut apart corresponding to M kind algorithm.
In the present embodiment, only with a kind of fuzzy clustering algorithm (fuzzy C-means clustering algorithm C for example
f), to select the situation of different initiation parameters be that example describes method of the present invention, but not as limitation of the present invention.
For the different fuzzy clustering algorithm of M kind, adopt following operation:
(1) sets this fuzzy C-means clustering algorithm C at random
fInitial parameter.Wherein, can different initiation parameters be set according to the characteristics of employing algorithm.If select the diverse fuzzy clustering algorithm of M kind, then do not need this step.
(2) after initiation parameter sets, use fuzzy C-means clustering algorithm C
fInput picture is cut apart.Each pixel in the image can be expressed as one five dimensional vector [R, G, B, x, y] and (also can be expressed as tri-vector [R as input, G, B]), R here, G, B represents redness, green, the blue component value of this pixel respectively, and x, y represent the position of this pixel, through fuzzy C-means clustering algorithm C
fAfter the processing, output degree of membership matrix [μ
j(x
i)]
K * N, μ wherein
j(x
i) i picture element belongs to the degree of membership (can be understood as the possibility that i picture element belongs to j classification) of j classification in the presentation video.
(3) repeat M above-mentioned steps (1), (2), promptly utilize the different fuzzy clustering algorithm of M kind to finish this image is cut apart, correspondingly obtain M degree of membership matrix [μ
j (1)(x
i)]
K * N, [μ
j (2)(x
i)]
K * N..., [μ
j (M)(x
i)]
K * N, correspond respectively to above-mentioned M kind algorithm and cut apart the cluster result that obtains.
Then, M degree of membership matrix [μ to obtaining
j (1)(x
i)]
K * N, [μ
j (2)(x
i)]
K * N..., [μ
j (M)(x
i)]
K * NCarry out registration.
Can adopt list of references 1: " Cluster Ensembles:A Knowledge Reuse Frameworkfor Combining Multiple Partitions[J] .Journal of Machine Learning Research; 2003,3 (3): 583-617 " in method carry out registration.Only introduce a kind of implementation method herein:
With first degree of membership matrix μ
j (1)(x
i)]
K * NThe classification mark as the benchmark of registration, establish its classification mark and be respectively C
1 (1), C
2 (1)... C
K (1)(picture element that promptly belongs to the k class is labeled as digital C
k (1)).If remaining other degree of membership matrixes [μ
j (2)(x
i)]
K * N..., [μ
j (M)(x
i)]
K * NIn, the classification of t degree of membership matrix is labeled as C
1 (t), C
2 (t)..., C
K (t), calculate wherein any one classification mark C
l (t), l=1,2 ..., K is with the classification mark C of the k class of first degree of membership matrix
k (1)The distance of corresponding degree of membership:
And calculate
C then
l (t)The classification mark to answer positioned in registration be C
Lopt (1)For example, suppose the d (1) that calculates, d (2) ... minimum value is d (x), then C among the d (K)
l (t)The classification mark should be set to C
x (1)
With the classification mark of each the degree of membership matrix in other degree of membership matrixes of remainder all registration be corresponding reference category mark, and with reference to registration results, with other degree of membership matrixes [μ of remainder
j (2)(x
i)]
K * N..., [μ
j (M)(x
i)]
K * NAll arrange, together with this benchmark matrix μ with reference to this benchmark matrix registration again
j (1)(x
i)]
K * NObtainable M the degree of membership matrix behind registration:
[μ
j (1)(x
i)]
K×N′,[μ
j (2)(x
i)]
K×N′,...[μ
j (M)(x
i)]
K×N′。
At last, with the M behind the registration degree of membership matrix [μ
j (1)(x
i)]
K * N', [μ
j (2)(x
i)]
K * N' ..., [μ
j (M)(x
i)]
K * N' be weighted fusion, obtain a weighting degree of membership matrix after the fusion, and according to this weighting degree of membership matrix computations classification mark of each picture element in the picture of publishing picture, thereby realize image segmentation.
Wherein, mainly comprise:
1) calculates the average mutual information Φ of the degree of membership matrix behind each registration according to following formula
t:
Wherein, M is the number of the fuzzy clustering algorithm of use; N is the number of pixels of image to be split; The classification number of K for setting; T=1,2 ..., M; μ
j (t)(x
iI the picture element of (t) individual matrix belongs to the degree of membership of j class in the degree of membership matrix behind collected M the registration of) ' expression; μ
j (l)(x
iI the picture element of (l) individual matrix belongs to the degree of membership of j class in the degree of membership matrix behind collected M the registration of) ' expression.
Φ
tValue big more, expression (t) individual degree of membership matrix is that comprised just few more with different information other degree of membership matrixes, then this degree of membership matrix is more little for the contribution of cluster fusion results.
2) calculate the weight of the degree of membership matrix behind each registration:
T=1 wherein, 2 ..., M; Z is a normalized factor, makes
3) the weighting degree of membership after calculating is merged:
I=1 wherein, 2 ..., N; J=1,2 ..., K; μ
j (t)(x
iI the picture element of (t) individual matrix belongs to the degree of membership of j class in the degree of membership matrix behind collected M the registration of) ' expression.
4) calculate each picture element x
iClassification, promptly for each i=1,2 ..., N, calculate:
I.e. each picture element x in this weighting degree of membership matrix
iPairing weighting degree of membership in the pairing classification j of maximal value be this picture element x
iThe classification mark.For example, suppose in this weighting degree of membership matrix picture element x
1The weighting degree of membership of a pairing K classification is respectively
Maximal value wherein is
, the subscript L of this maximal value correspondence is exactly pixel x
1The classification mark.
After finishing this step, each picture element in the input picture has all obtained its corresponding class mark, has also just finished image segmentation.If desired, can also carry out aftertreatment, such as removing noise spot (as isolated point), merging the very little zone of area to cluster result, the meaning of aftertreatment is, consider the needs of practical application, some apparent errors in the segmentation result of place to go make segmentation result more reasonable.
Compared with prior art, remarkable advantage of the present invention is: adopt fuzzy technology to handle color digital image, and merge the segmentation result of multiple fuzzy clustering algorithm, further improved the precision of color images.
Certainly; the present invention also can have other various embodiments; under the situation that does not deviate from spirit of the present invention and essence thereof; those of ordinary skill in the art work as can make various corresponding changes and modification according to the present invention, but these corresponding changes and distortion all should belong to the protection domain of the appended claim of the present invention.
Claims (12)
1. a color image segmentation method is characterized in that, comprising:
Import image to be split and carry out initializing set, set the classification number of image to be split;
Use multiple different fuzzy clustering algorithm respectively this image to be cut apart, correspondingly obtain a plurality of degree of membership matrixes;
With in these a plurality of degree of membership matrixes one of them as the benchmark matrix, classification mark to other degree of membership matrixes of remainder carries out registration respectively, and according to the classification mark behind the registration each degree of membership matrix is carried out again registration and arrange the corresponding degree of membership matrix that obtains behind registration;
Described degree of membership matrix and this benchmark matrix behind registration merged, obtain a degree of membership matrix after the fusion, and go out the pairing classification mark of each picture element, thereby realize image segmentation according to the degree of membership matrix computations after this fusion.
2. color image segmentation method according to claim 1 is characterized in that, the multiple different fuzzy clustering algorithm of described use is cut apart this image, and the step that obtains a plurality of degree of membership matrixes comprises:
Set a kind of initial parameter of fuzzy clustering algorithm;
Utilize this fuzzy clustering algorithm that this image is cut apart, obtain a degree of membership matrix;
Repeat repeatedly above-mentioned steps, utilize multiple different fuzzy clustering algorithm respectively this image to be cut apart, correspondingly obtain a plurality of degree of membership matrixes.
3. color image segmentation method according to claim 2 is characterized in that, described step of carrying out registration comprises:
Select in these a plurality of degree of membership matrixes one of them as the benchmark matrix, set its classification mark and with as the reference category mark, wherein, the picture element cluster that belongs to same classification is same classification mark;
Other degree of membership matrixes for remainder, set the classification mark of each degree of membership matrix respectively, and calculate the distance of the pairing degree of membership of each reference category mark of any one classification mark and this benchmark matrix in the classification mark of each degree of membership matrix, then to answer positioned in registration be the minimum reference category mark of distance to this any one classification mark;
With the classification mark of each the degree of membership matrix in other degree of membership matrixes of remainder all registration be corresponding reference category mark, and with reference to registration results, other degree of membership matrixes of remainder are carried out again registration arrangement, a plurality of degree of membership matrixes behind registration of corresponding acquisition respectively.
4. color image segmentation method according to claim 3 is characterized in that, the described step that merges and calculate each picture element corresponding class mark comprises:
Collect this benchmark matrix and a plurality of degree of membership matrixes behind registration;
Calculate the average mutual information of each matrix in this benchmark matrix and a plurality of degree of membership matrixes behind registration;
Calculate the weight of each matrix;
This benchmark matrix and a plurality of degree of membership matrixes behind registration are weighted fusion, obtain a weighting degree of membership matrix after the fusion;
According to the weighting degree of membership matrix after this fusion, calculate the classification mark of each picture element in the fused image, finish image segmentation.
5. color image segmentation method according to claim 4 is characterized in that, in the described step of carrying out registration, the distance of this degree of membership satisfies:
Wherein, N is the number of pixels of image to be split; K, l represent classification; μ
k (1)(x
i) represent in this benchmark matrix (1) that i picture element belongs to the degree of membership of k class; μ
l (t)(x
i) represent that i the picture element of (t) individual matrix in other remaining degree of membership matrixes belongs to the degree of membership of l class.
6. color image segmentation method according to claim 5, it is characterized in that, in the described step of carrying out registration, for other degree of membership matrixes of remainder, the classification mark of any one classification mark behind registration satisfies in the classification mark of each degree of membership matrix:
7. according to the color image segmentation method shown in the claim 6, it is characterized in that described average mutual information Φ
tSatisfy:
Wherein, M is the number of the fuzzy clustering algorithm of use; N is the number of pixels of image to be split; The classification number of K for setting; μ
j (t)(x
iI the picture element of (t) individual matrix belongs to the degree of membership of j class in the degree of membership matrix behind collected M the registration of) ' expression; μ
j (l)(x
iI the picture element of (l) individual matrix belongs to the degree of membership of j class in the degree of membership matrix behind collected M the registration of) ' expression.
8. color image segmentation method according to claim 7 is characterized in that, the weight of the degree of membership matrix behind each registration satisfies:
T=1 wherein, 2 ..., M; Z is a normalized factor, makes
9. color image segmentation method according to claim 8 is characterized in that, described weighting degree of membership matrix satisfies:
I=1 wherein, 2 ..., N; J=1,2 ..., K; μ
j (t)(x
iI the picture element of (t) individual matrix belongs to the degree of membership of j class in the degree of membership matrix behind collected M the registration of) ' expression.
10. color image segmentation method according to claim 9 is characterized in that, the classification mark of described each picture element satisfies:
Each picture element x in this weighting degree of membership matrix wherein
iPairing weighting degree of membership in the pairing classification j of maximal value be this picture element x
iThe classification mark.
11. according to the described color image segmentation method of above-mentioned arbitrary claim, it is characterized in that, also comprise before the image segmentation carrying out:
The image of being imported to be split is carried out pre-service, remove the noise spot in the image.
12. color image segmentation method according to claim 11 is characterized in that, also comprises after finishing image segmentation:
Described image segmentation result is carried out aftertreatment, remove the very little zone of area in noise spot in the image and/or the combined diagram picture.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2008100558332A CN101216890B (en) | 2008-01-09 | 2008-01-09 | A color image segmentation method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2008100558332A CN101216890B (en) | 2008-01-09 | 2008-01-09 | A color image segmentation method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN101216890A true CN101216890A (en) | 2008-07-09 |
CN101216890B CN101216890B (en) | 2011-02-16 |
Family
ID=39623321
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2008100558332A Expired - Fee Related CN101216890B (en) | 2008-01-09 | 2008-01-09 | A color image segmentation method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN101216890B (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101853494A (en) * | 2010-05-24 | 2010-10-06 | 淮阴工学院 | Color image segmentation method based on coring fuzzy Fisher criterion clustering |
CN101976438A (en) * | 2010-10-27 | 2011-02-16 | 西安电子科技大学 | FCM (Fuzzy Cognitive Map) texture image segmentation method based on spatial neighborhood information |
CN102737381A (en) * | 2012-06-13 | 2012-10-17 | 西安电子科技大学 | Image partitioning method based on mixed bipartite graph clustering integration |
CN102881019A (en) * | 2012-10-08 | 2013-01-16 | 江南大学 | Fuzzy clustering image segmenting method with transfer learning function |
WO2019238126A1 (en) * | 2018-06-15 | 2019-12-19 | 北京市商汤科技开发有限公司 | Image segmentation and segmentation network training method and apparatus, device, medium, and product |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
GB2395263A (en) * | 2002-11-12 | 2004-05-19 | Qinetiq Ltd | Image analysis |
CN1238816C (en) * | 2003-03-14 | 2006-01-25 | 清华大学 | Font recogtnizing method based on single Chinese characters |
CN1252588C (en) * | 2003-05-16 | 2006-04-19 | 上海交通大学 | High spectrum remote sensing image combined weighting random sorting method |
CN1256704C (en) * | 2003-10-30 | 2006-05-17 | 上海交通大学 | Method for picking up and comparing spectral features in remote images |
JP4561273B2 (en) * | 2004-09-22 | 2010-10-13 | ソニー株式会社 | Image processing apparatus and method, and program |
-
2008
- 2008-01-09 CN CN2008100558332A patent/CN101216890B/en not_active Expired - Fee Related
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101853494A (en) * | 2010-05-24 | 2010-10-06 | 淮阴工学院 | Color image segmentation method based on coring fuzzy Fisher criterion clustering |
CN101853494B (en) * | 2010-05-24 | 2011-11-30 | 淮阴工学院 | Color image segmentation method based on coring fuzzy Fisher criterion clustering |
CN101976438A (en) * | 2010-10-27 | 2011-02-16 | 西安电子科技大学 | FCM (Fuzzy Cognitive Map) texture image segmentation method based on spatial neighborhood information |
CN101976438B (en) * | 2010-10-27 | 2012-03-28 | 西安电子科技大学 | FCM (Fuzzy Cognitive Map) texture image segmentation method based on spatial neighborhood information |
CN102737381A (en) * | 2012-06-13 | 2012-10-17 | 西安电子科技大学 | Image partitioning method based on mixed bipartite graph clustering integration |
CN102737381B (en) * | 2012-06-13 | 2014-10-29 | 西安电子科技大学 | Image partitioning method based on mixed bipartite graph clustering integration |
CN102881019A (en) * | 2012-10-08 | 2013-01-16 | 江南大学 | Fuzzy clustering image segmenting method with transfer learning function |
CN102881019B (en) * | 2012-10-08 | 2014-11-19 | 江南大学 | Fuzzy clustering image segmenting method with transfer learning function |
WO2019238126A1 (en) * | 2018-06-15 | 2019-12-19 | 北京市商汤科技开发有限公司 | Image segmentation and segmentation network training method and apparatus, device, medium, and product |
Also Published As
Publication number | Publication date |
---|---|
CN101216890B (en) | 2011-02-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Cai et al. | A unified multi-scale deep convolutional neural network for fast object detection | |
CN104809187B (en) | A kind of indoor scene semanteme marking method based on RGB D data | |
CN107704877A (en) | A kind of image privacy cognitive method based on deep learning | |
Arbelle et al. | Microscopy cell segmentation via adversarial neural networks | |
CN101216890B (en) | A color image segmentation method | |
CN101840577B (en) | Image automatic segmentation method based on graph cut | |
CN101169867B (en) | Image dividing method, image processing apparatus and system | |
CN104537126B (en) | A kind of overlapping community discovery method based on edge graph random walk | |
CN102165491A (en) | Image processing device, image processing method, and program | |
CN113052254B (en) | Multi-attention ghost residual fusion classification model and classification method thereof | |
Ros et al. | Unsupervised image transformation for outdoor semantic labelling | |
Jdid et al. | Robust automatic modulation recognition through joint contribution of hand-crafted and contextual features | |
CN109685772B (en) | No-reference stereo image quality evaluation method based on registration distortion representation | |
CN104282022A (en) | Fuzzy clustering image segmenting method with multi-view-angle learning ability | |
Rios et al. | Feature visualization for 3D point cloud autoencoders | |
CN101216891B (en) | A color image segmentation method | |
CN106874931B (en) | User portrait clustering method and device | |
CN111353988A (en) | KNN dynamic self-adaptive double-image convolution image segmentation method and system | |
CN110929744B (en) | Hierarchical joint convolution network feature-based weak supervision image semantic segmentation method | |
Kumar et al. | A hybrid cluster technique for improving the efficiency of colour image segmentation | |
CN112802048B (en) | Method and device for generating layer generation countermeasure network with asymmetric structure | |
KR20230073751A (en) | System and method for generating images of the same style based on layout | |
CN108876803A (en) | A kind of color image segmentation method based on spectral clustering community division | |
Yao et al. | AGO: accelerating global optimization for accurate stereo matching | |
CN103942779A (en) | Image segmentation method based on combination of graph theory and semi-supervised learning |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20110216 Termination date: 20200109 |
|
CF01 | Termination of patent right due to non-payment of annual fee |